Fix #448: switch to Type=oneshot (#449)

With Type=forking systemd tried to monitor the forked process. This
allows it to know things like when the service dies. But, the way
vagrant (and vbox underneath) daemonizes confuses systemd. So, if
there are multiple vm's started that are managed by systemd all but
the first start and then are immediately shutdown.

Type=oneshot makes it so systemd starts the process but then doesn't
monitor it. RemainAfterExit=yes makes it so systemd assumes the
service stays running. KillMode=none makes it so on systemctl stop
systemd doesn't send a kill to all processes that were started when
systemctl start was executed. On start of first vm on the machine
VirtualBox starts a server daemon process which needs to not be
killed.

These settings make it so systemd states will only change when a
systemctl command is issued. This can lead to confusing behavior where
a vm might stop outside of systemd's control (ex a manual vagrant
halt) but systemd will still report the service as "active".

rsconf/package_data/vm_devbox/vm_devbox_unit_service.jinja

@@ -6,9 +6,11 @@ Environment=TZ=:/etc/localtime
 ExecStart={{ this.start_f }}
 ExecStop={{ this.stop_f }}
 Group={{ this.run_u }}
+KillMode=none
+RemainAfterExit=yes
 SyslogIdentifier={{ systemd.service_name }}
 TimeoutStartSec={{ this.timeout_start_min }}min
-Type=forking
+Type=oneshot
 User={{ this.run_u }}
 
 [Install]